Audibility-index functions for the connected speech test

Abstract

Objective: A study was performed to derive a frequency importance function (FIF), a performance-intensity funcTIon (PIF), and a transfer function (TF) for the audio compact disc version of the Connected Speech Test (CST).

Design: CST passages were masked with talker-spectrum-matched (TSM) noise and presented to two groups of normal-hearing adult subjects. One group (N = 48) listened through a wideband filter, seven low-pass filters, and eight high-pass filters at six signal-to-noise (S/N) ratios. The other group (N = 12) listened through just the wideband filter at 12 S/N ratios. The FIF was based on the data for the first group while the PIF was based on the data for the second group. The results of both groups were used to determine the TF between CST scores and audibility-index (AI) values.

Results: The FIF, in 1/3-octave bands, is a bimodal curve with a minor peak at 500 Hz, a major peak at 1600 to 2000 Hz, and a midpoint of 1619 Hz. The PIF in TSM noise and the TF are both asymmetric S-shaped curves; their respective slopes are 12%/dB and 11%/0.0333 AI.

Conclusions: Comparisons between these results and the findings of other studies reconfirm that different speech materials have different AI functions. The FIF for the CST overlaps 1/3-octave band functions for continuous discourse and average speech but does not have the same shape as those functions. The TF indicates that CST passages are generally more intelligible than isolated monosyllabic words (NU6 lists) and somewhat less intelligible than continuous discourse. The former result is probably due at least partly to the effects of context whereas the latter result may be due primarily to how clearly the talkers pronounced the speech materials.